Mathematical Biology Seminar Scott Diamond, University of Pennsylvania Wednesday, February 7, 2024 3:05pm in LCB 222 Title: Patient-specific blood phenotyping to predict 3D coronary thrombosis with single platelet resolution Abstract: The response of flowing blood to ruptured coronary plaque can be life threatening. Thrombus growth is a complex and multiscale process involving interactions spanning length scales from platelets to the macroscopic clot. We developed pairwise agonist scanning (PAS) using 384-well plate robotic assay to generate high-dimensional dynamic data for neural network training of patient-specific platelet activation. Next, we implemented a multiscale framework to simulate thrombus growth under flow comprising four individually parallelized and coupled modules: the data-driven Neural Network (NN) that accounts for platelet calcium signaling, a LaNce Kinetic Monte Carlo (LKMC) simulation for tracking platelet positions, a Finite Volume Method (FVM) simulator for solving convection-diffusion-reaction equations describing agonist release and transport, and a LaNce Boltzmann (LB) flow solver for computing the blood flow field over the growing thrombus. Parallelization was achieved by developing in-house parallel routines for NN and LKMC, while the open-source libraries OpenFOAM and Palabos were used for FVM and LB respectively. Importantly, the parallel LKMC solver utilized particle-based parallel decomposition allowing efficient use of processors over highly varying regions of the problem. The modularity of our approach is a noteworthy feature, allowing for easy integration of new components in complex geometries as research progresses. This adaptability enhances its utility in studying diverse physiological scenarios and investigating the impact of various factors such as flow conditions, patient-specific vascular geometry, and drug treatments on thrombus growth.